Feasibility Study of Variance Reduction in the Logistics Composite Model

The Logistics Composite Model (LCOM) is a stochastic, discrete-event simulation that relies on probabilities and random number generators to model scenarios in a maintenance unit and estimate optimal manpower levels through an iterative process. Models such as LCOM involving pseudo-random numbers inevitably have a variance associated with the output of the model for each run, and the output is actually a range of estimates. The reduction of the variance in the results of the model can be costly in the form of time for multiple replications. The alternative is a range of estimates that is too wide to realistically apply to real-world maintenance units. This research explores the application of three different methods for reducing the variance of the model's output. The methods include Common Random Numbers (CRN), Control Variates, and Antithetic Variates. The differences in the 95% confidence intervals were compared between the variance reduction techniques and the original model to determine the degree of variance reduction. The result is a successful variance reduction in the primary output statistics of interest using the application of the Control Variates technique, as well as a methodology for the implementation of Control Variates in LCOM.